I think we all know the basics: stealing is great. But let's dig a bit deeper. What about the math, the EV stuff, and what about the CO?

In my 50nl 6m games at Full Tilt, well over half the players are TAGG-regs who THINK they can play exploitively, 3b light correctly and make the proper adjustments. They're 18/16/3 types who flat call about 8% of the time and with 6% 3b stats that mean they resteal about 10% of the time they're in the blinds and a steal-happy BTN raises.

Since they 3bet more vs. steals and tend to call wider, as well, let's assume that Hero is BTN and that the SB and BB both flat call with 10% of their hands and 3bet with another 10%.

BTN Steal Success

So Hero wins 1.5 BB any time both blinds fold. Both blinds are TAGG's who by assumption will:

80% Fold
10% Call
10% 3bet

So they both fold 64% of the time ( 0.8 * 0.8 ). One or the other 3bet's 18% of the time ( 0.1 + 0.8 * 0.1 = 0.1 + 0.08 ). That leaves some squeezes we haven't accounted for (say 2%) with the rest being 1 or 2 callers (say 16%).

Assume Hero ditches to any squeeze or 3bet, and loses 3 BB's. For the moment, let's assume Hero breaks even when called since he has position, even though he'll most often has the worse preflop hand. Then the EV calculation is:

EV(both fold) = 0.64 * 1.5 BB = 0.96 BB
EV(3b/sqz) = 0.2 * (-3.0 BB) = -0.6 BB
EV(call[s]) = 0.16 * (?? BB) = 0 BB (by assumption)

So Hero makes 0.96 - 0.60 = 0.36 BB profit for every steal attempt from the BTN (36 bb/100 winrate). But then the question arises: what is the likely EV of trying to steal with crap cards and getting called in at least one spot?

I used the HEM "steal attempted" filter to analyze my last 9k 50n 6m HH's. I attempted to steal 1330 times, 1001 of which were from the CO or BTN (SB steals are a bit different dynamic because Hero is oop any time the steal isn't successful).

For those 1001 hands, I won $409.03 (81.8 bb/100 winrate).

Of those 1001 hands, I saw the flop 327 times. (Used HEM "more filters" tab: SawFlop = True). Remember that the "folds to 3bet" losses are included above, so these are just the times I had to play a pot postflop after being the first to raise in the CO or BTN.

For those 327 hands, I won $178.30 (109.05 bb/100).

Interestingly, the winrate was higher when I got called than when I was cbetting. Of those 327 hands, I filtered for when a cbet was possible, which turned out to be 229 times.

For those 229 hands, I won $23.35 (20.39 bb/100).

Again, it's interesting that the times when villain donk-bet the flop showed a HIGHER winrate than when he checked to me. (Check raises are often good to bluff or semi-bluff with if you're in the blinds here and being stolen from, btw!!) These samples are smallish, so I don't want to act like this is ironclad. But it is interesting. FWIW, I also filtered on HU flops vs. 3-way flops, but the winrate in bb/100 is roughly the same.

The whole point is that the red ?? in the steal EV formula likely represent a positive number and mean we can steal more often and against less TAGGish villains.


CO Steal Success

The next thought that occurred to me was this: if I can be profitable when I'm "caught stealing" (but flatted, not raised), then what are the EV implications for the CO? Well, I reran the filters to get this breakdown: 482 CO steal attempts.

For those 482 hands, I won $83.25 (34.54 bb/100).

Of those, I saw the flop 167 times, with $25.50 lost (-30.54 bb/100). I was able to cbet 123 times, and lost $43.70 (-71.06 bb/100) on those.

It should be obvious the difference is due to the BTN jumping in the way. So if we're stealing from the CO, we need not just a TAGG, but a nit, someone who will fold his BTN a huge majority of the time. But after that, we're golden since the CO steal is much less often 3b against. When you steal from the BTN, it's obvious what you're doing, right? When you steal from the CO, if the villains in the blinds aren't checking popup stats, they think you might have a hand. So they call and 3bet less often. Let's say they flat and 3bet 8% of the time each (they probably don't do it quite this much, especially 3b, but it will keep our EV estimates conservative).

Then, the probability of the blinds folding/calling/3betting (given that the BTN has already) can be recalculated:

Both fold: 71% of the time ( 0.84 * 0.84 )
At least one 3bet's: 15% of the time ( 0.08 + 0.84 * 0.08.
Squeezes (say 3%) with the rest being 1 or 2 callers (~ 11%).

As long as the BTN folds, the steal has the same EV against the blinds. So let's recalculate the EV of a BTN steal with these percentages so that we have an idea of how often we need the BTN to fold so that we can show a profit.

EV(both fold) = 0.71 * 1.5 BB = 1.06 BB
EV(3b/sqz) = 0.18 * (-3.0 BB) = -0.44 BB
EV(call[s]) > 0 (by analysis above).

So the EV of a CO steal when the BTN folds is now 0.62 BB, or at least 62 bb/100.

Suppose that Hero just folds postflop if the BTN calls and no one raises. Then Hero loses 3.5 BB any time the BTN 3bets or calls and wins .62 BB (his steal EV against the blinds) any time the BTN folds. Let F = BTN fold percentage, and then:

Overall EV (CO steal) = F * 0.62 - ( 1 - F ) * 3.5 = 4.12 F - 3,

which is positive provided F > .85. So if Hero can expect the BTN to fold at least 85% of the time, he can steal WITH ANY TWO CARDS AND SHOW A PROFIT. Of course, I don't actually recommend stealing with any two - even the most obtuse villains will catch on to that. But you can steal with a pretty wide range, say 40% or more.

Stealing in Practice

The key to "stealing" effectively obviously hinges on reading when opponents are unlikely to call or 3bet and then expanding our range to many (but not really any two). And stealing from the CO is a good idea especially with a TAGG BTN, since they don't tend call or 3b wide enough to dent our profits.

The biggest difficulty with stealing comes from the marginal hands played badly, so you have to have the discipline to fold to lots of 3bets. If you try a steal with K2s and get 3bet by a TAGG BB, you've gotta ditch. NH to him, especially if he's 3betting light. He made the right play, but he's probably (if he's like the TAGG's I play against) not capable of making a big enough adjustment to deny you profits from steals. Where he'll kill your profits is the big hands you should have folded either preflop or on the flop, but that you couldn't get away from.

So if we're going to try stealing 35-40% or more of the time in the CO/BTN (which is probably a good idea), you've got to get away from hands where we're really only likely to win small pots and lose bigger ones.

By the way, for anyone interested, my stealing rate in the above 9k hand sample was 49%, or 41% CO, 56% BTN, and 59% SB for 1330 total steal attempts, with a "steal success" rate of 59%. Overall, I was up $460 from steal attempts (69.2 bb/100). That was $50 in the SB (31 bb/100), $83 in the CO (34.5 bb/100) and $326 from the BTN (125.7 bb/100).

When I filtered, I found I was folding to more than 75% of resteals (3bets) I faced. When I called, I basically broke even (+$4 over 41 hands). When I 4b, I won nicely ($75 over 15 hands). So overall resteals suck, but I was picking the right places to play back.

Just some stuff I was working on lately and thought I'd share. Be interested to hear your thoughts. I would also be interested to hear from someone who has played the same stealing ranges over a larger sample at the same stakes. I may dig deeper into my past HH's if this thread goes anywhere.

That's enough for now - jeez I just realized it's 2:45 AM and I have to be at work tomorrow.

Good luck at the tables.